Semantic Transparency and Cognitive Modelling of Complex Performance

Abstract

The simulation of human performance in situations involving interaction with complex technological interfaces highlights some key issues in cognitive modelling. This is illustrated in the behaviour of a pilot engaged in air combat. Modelling a pilot’s situation awareness requires an approach that provides for emotional and motivational processes as weil as sensori-motor and cognitive processes. This strategic decision raises the question of the type of representation to be employed. A response involves determining the relative roles of symbolic and sub-symbolic representation and, more specifically, the relative merits of rules and neural nets.The concept of a spectrum of semantic transparency is introduced as a basis for the theoretical analysis of processes underlying human performance in terms relevant to ANN and symbolic representations. This leads to recognition of the inevitability of employing a variety of hybrid forms of representation. It, also, highlights the need for exploration of architectural frameworks to identify those that facilitate articulation and integration of forms of representation and realisation of the potential benefits of their complementarity.A novel hybrid cognitive architecture is presented which employs interaction between symbolic and sub-symbolic networks to represent the effects of emotional and motivational processes on decision making. This enables a shift from the conventional focus of combat simulation on optimal decision making to the modelling of authentic human decision making with its individual variability and imperfection. The potential of the architecture and hybrid form of representation to resolve general difficulties encountered in the simulation of human performance is discussed.KeywordsSituation AwarenessSymbolic RepresentationSemantic NetworkConscious AwarenessMotivational NetworkThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.